Phase 3: Automation Engine

Status: ⏳ In Development

Multi-stage pipeline transforming requests into downloaded, organized media in Plex.

Pipeline

Request → search_indexers → rank_results → download_torrent
→ monitor_download → process_audiobook → update_plex

Job Types

search_indexers - Search Prowlarr for torrents
rank_results - Apply ranking algorithm, select best
download_torrent - Add to download client (qBittorrent/Transmission/SABnzbd)
monitor_download - Poll progress (10s intervals)
process_audiobook - Organize files to media directory
update_plex - Trigger scan, fuzzy match

Automatic searches keep the ranked candidate list. If Prowlarr returns HTTP 429 or 5xx while grabbing the selected release, download_torrent attributes the error to the indexer/source and falls through to the next release. When all candidates have temporary grab failures, Bull retries the job with backoff.

Integration Points

Indexers: Prowlarr (primary), Jackett (fallback) Download Clients: qBittorrent or Transmission (torrent), SABnzbd (usenet) — details Media Server: Plex (scan + match)

Job Queue (Bull)

Redis-backed for persistence
Retry: 3 attempts, exponential backoff (2s, 4s, 8s)
Priorities: High (10), Medium (5), Low (1)
Concurrency: 3 concurrent per type
Jobs survive app restarts

Config Keys

Prowlarr: indexer.type=prowlarr, indexer.prowlarr_url, indexer.prowlarr_api_key qBittorrent: download_client.type=qbittorrent, download_client.qbittorrent_url/username/password Paths: paths.download_dir, paths.media_dir

Prowlarr
Download Clients - Multi-client management, protocol routing
qBittorrent
SABnzbd
Ranking Algorithm
File Organization
Plex Integration

README.md